TCP/IP & the End-to-End Principle
Vint Cerf & Bob Kahn · 1974
"Cerf and Kahn designed the internet to be dumb at the core and smart at the edges — reliability and intelligence live in endpoints, not in the network — which is why the same packet fabric could absorb every application since without redesign."
In 1974 Cerf and Kahn made a contrarian bet: don't make the network reliable — make it simple and let the ends retry. That dumb core is why the same internet carries email in 1982 and video calls in 2024.
End-to-end argues that functions placed at low levels are redundant unless needed by all applications — so keep the core that moves packets minimal, and push error correction, encryption, and ordering to hosts that know what the application actually needs. The result is hourglass architecture: IP as narrow waist, everything above and below can evolve independently. Centralized 'smart network' alternatives (X.25, ATM) lost because they optimized for the current application and froze.
What was the architectural bet behind the end-to-end principle in TCP/IP?
Read more about the topic
The explanation above is written with AI assistance. These are the originals — go to them to check it.
- A Protocol for Packet Network Intercommunication — TCP/IP (Cerf & Kahn, 1974)Vint Cerf & Bob Kahn / Wikipedia
The Unix Philosophy
"Unix won not by features but by composition — small tools that do one thing well, speak in text streams, and compose via pipes, so new capabilities emerge from combining old parts rather than rewriting."